{
  "schema_version": 1,
  "builder_plan": {
    "scenario": "An agent looks good in demos, but you cannot tell which product tasks it fails.",
    "change": "Use the study's evaluation idea to inspect a small set of tasks with human-checked answers.",
    "baseline": "Your current evaluator and a manually reviewed reference set.",
    "steps": [
      "Locate the paper's implementation and check its license, model support, data needs, and hardware requirements. Stop here if you cannot obtain the required artifacts.",
      "Save a small set of representative product tasks and expected outcomes. Keep evaluation tasks out of any training or tuning, and choose quality and cost limits before running.",
      "Run the baseline and the proposed change on identical inputs. Keep other settings fixed and repeat paired runs to expose variation.",
      "Save per-task outputs, failures, timings, and all usage costs. Review regressions before expanding the test."
    ],
    "measure": "Agreement with human review, missed failures, false alarms, and evaluation cost per task.",
    "decision": "Keep the change only if it meets your preselected quality and latency limits and improves the metric you care about after setup and operating costs. Otherwise retain the baseline. This is an exploration plan, not a validated recommendation."
  },
  "builder_review": null,
  "arxiv_id": "2610.31148",
  "title": "Prompt Sensitivity in Tool-Augmented Model Comparisons",
  "summary": "This replaces single-wording, accuracy-only benchmarking of tool-augmented models with a matched evaluation protocol that separates answer readout, reasoning, the call decision, the tool's answer, and the no-call report, then measures pairwise comparison stability across paraphrases. It costs additional evaluation runs (five wordings per question), call-frequency instrumentation, and more complex analysis, but no extra training or hardware beyond existing inference. After adoption, stability conclusions can still depend on the tool-calling interface and instructions: a native interface and alternate tool descriptions removed detected widening in one domain, and sensitivity persists wherever tools return intermediate results, so deterministic tool outputs alone do not stabilize model rankings.",
  "rationale": "The reader already compares models for production selection, and this protocol requires only repeated inference with reworded prompts plus call logging, which fits a single 24 GB GPU or API budget and directly reduces the risk of choosing a model whose benchmark lead is an artifact of prompt wording.",
  "technique": "Prompt-sensitivity protocol for tool-use model comparison",
  "family": "avaliacao_de_agente",
  "recommendation": "adotar",
  "published": "2026-10-22",
  "as_of": "2026-10-23",
  "editorial_status": "indexed",
  "source_url": "https://arxiv.org/abs/2610.31148",
  "independent_implementations": 1,
  "total_implementations": 2,
  "stars": 122,
  "citations": null,
  "claims": [
    {
      "claim_id": "claim-01",
      "statement": "This replaces single-wording, accuracy-only benchmarking of tool-augmented models with a matched evaluation protocol that separates answer readout, reasoning, the call decision, the tool's answer, and the no-call report, then measures pairwise comparison stability across paraphrases. It costs additional evaluation runs (five wordings per question), call-frequency instrumentation, and more complex analysis, but no extra training or hardware beyond existing inference. After adoption, stability conclusions can still depend on the tool-calling interface and instructions: a native interface and alternate tool descriptions removed detected widening in one domain, and sensitivity persists wherever tools return intermediate results, so deterministic tool outputs alone do not stabilize model rankings.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2610.31148",
      "source_page": null,
      "source_excerpt": ""
    }
  ],
  "exposure_map": [
    {
      "dimension": "quality",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "compute",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "latency",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "operations",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "compatibility",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "security",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "data_and_training",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "reproducibility",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    }
  ],
  "risks": [],
  "minimum_test": [],
  "open_questions": [],
  "independent_tests": [],
  "lab_reproductions": [],
  "report_available": false,
  "validation_tier": "unknown",
  "evidence_tier": "unknown",
  "infrastructure_basis": "unknown",
  "equations": [],
  "equations_status": "not_fetched",
  "core_kind": "",
  "equations_fetched_at": "",
  "equations_source_url": "",
  "equations_source_sha256": "",
  "exposures_reviewed_at": "",
  "exposures_source_sha256": ""
}